E-bike impacts: quantifying the energy use and lifecycle emissions in response to real world driving conditions

(1)

E-bike Impacts: Quantifying the Energy Use and Lifecycle Emissions in Response to Real World Driving Conditions

by

Daniel Clancy

B.Eng, University of Victoria, 2015

A Thesis Submitted in Partial Fulfillment of the Requirements for the Degree of

MASTER OF APPLIED SCIENCE

in the Department of Mechanical Engineering

(2)

Supervisory Committee

E-bike Impacts: Quantifying the Energy Use and Lifecycle Emissions in Response to Real World Driving Conditions

by

Daniel Clancy

B.Eng, University of Victoria, 2015

Supervisory Committee

Dr. Curran Crawford, Department of Mechanical Engineering Co-Supervisor

Dr.Nedjib Djilali, Department of Mechanical Engineering Co-Supervisor

(3)

Abstract

E-bikes can significantly enhance adoption of cycling as an urban transportation mode and have the advantage of low space requirements, very small operational GHG

emissions, and a negligible contribution to infrastructure degradation. This work explores some key environmental and physical performance features of E-bikes in real world settings in order to systematically determine the capabilities of E-bikes for greater adoption. This includes analysis on the lifecycle emissions associated with E-bikes and comparisons with other major urban modes. Empirical data was collected about the performance of a third-party electric motor technology that could improve energy efficiency. The ability of this technology to offer improvements under real world conditions was verified and showed promise with recommendations for further

development. A trial of E-bikes deployed in a corporate fleet, with 17 riders and over 600 km of trip data was completed and used for several additional analyses. This include validating a mathematical model of an E-bike, as well as extending the boundaries of previous lifecycle research to include upstream dietary emissions associated with human supplied mechanical power while riding an E-bike. The results in this thesis show both the strengths of E-bikes as used for corporate and personal transportation, as well as the barriers that still remain for greater adoption.

(4)

List of Tables

Table 1: Estimated land area requirements of various modes of transport [6]. ... 6

Table 2: Embodied Energy of various transport modes divided among passenger load [19] ... 8

Table 3: Summary of Initial Exro Project Test Equipment ... 14

Table 4: Summary of Final Expo Project Test Equipment ... 15

Table 5: Quality of fit parameters for surface fit to first testing method data. ... 20

Table 6: Quality of fit parameters for surface fit to second testing method data. ... 24

Table 7: CRD Project data collection summary. ... 28

Table 8: Average energy and power for total recorded data ... 29

Table 9: Dietary Emissions from a range of UK and USA diet types [46], [47] ... 38

Table 10:Summary of electrical grid emission intensities for UK and USA for 2017 [48], [49] ... 39

Table 11: Energy use of bicycles in urban commuting [52], [53]. Rider EE reported as human caloric expenditure, other energy values reported as output at pedals. ... 41

Table 12: E-bike system efficiency estimates ... 42

Table 13: E-bike primary energy intensity as determined from CRD project data. ... 43

Table 14: Chung method drag area and rolling resistance values. ... 57

Table 15: Model constant values along with source and measured or estimated uncertainty... 57

Table 16: Simply and Dynamic trip power prediction uncertainty contributions from each variable. Mean power uncertainty (Mean Ci) as well as percentage of average predicted power (Ci/P0) shown. ... 63

Table 17: Comparison of model prediction error for Simple trip with and without grade, dynamic trip, and the results from Dahmen et al’s work. ... 64

Table 18: Well-to-wheel emission intensity along with typical urban pace for E-bikes, fossil fuel and electric cars, and walking. Values marked with an asterisk (*) come from Weiss et al’s 2015 study. Pace includes average of all recorded data, including zero speeds [41], [72], [73]. ... 75

Table 19: CRD Project well-to-wheel emissions for E-bikes compared to CRD ICE and BEV sedans. A total of 607 km were logged during the CRD project. ... 76

Table 20: E-bike, fossil fuel and electric car capital and operational costs per vehicle representing ownership over 5 years [74]–[76]. ... 77

Table 21: Modified model input variables for assessing E-bike cargo performance. ... 79

(7)

List of Figures

Figure 1: On the left, a typical Mid-Drive, geared motor E-bike, the Norco VLT as used in CRD E-bike Trial. On the right, a hub-located direct-drive e-bike. Images from

www.norco.com and www.publicbikes.com ... 3 Figure 2: Example of initial test results for a fixed braking force as the Bionx system is accelerated from rest to maximum speed. ... 17 Figure 3: Entire range of achieved torque-RPM states for two rounds of testing using the first method and equipment. A) first round of testing, b) second round of testing ... 18 Figure 4: Polynomial surface fits to A) round 1 and B) round 2 state data of first testing method. Contours show system efficiency. ... 19 Figure 5: Absolute difference between the two efficiency surface fits of Figure 4 ... 19 Figure 6: Distribution of residuals for surface fit to a) round 1 and b) round 2 data using the first testing method. Mean and standard deviation of error from surface fit shown for each round of testing. ... 20 Figure 7: Achieved torque-RPM states for two rounds of testing using the second method for both parallel and series wiring configurations. a) and b) are parallel tests, c) and d) are series tests. ... 22 Figure 8: Polynomial surface fits to second method results. a) and b) use parallel test data, c) and d) use series test data. ... 23 Figure 9: Absolute difference between round and 2 efficiency maps for second testing method. a) parallel and b) series wiring configurations. ... 24 Figure 10: Distribution of residuals for surface fit to round 1 and 2 results using second testing method. a) Parallel round 1, b) parallel round 2, c) series round 1, d) series round 2... 25 Figure 11: Efficiency maps for both rounds combined. a) parallel combined efficiency map, b) series combined efficiency map, c) parallel individual efficiency data points, d) series individual efficiency data points. ... 26 Figure 12: Per-trip energy use (top), average per-trip power (middle), and trip distance (bottom), as recorded during the CRD project. ... 30 Figure 13: E-bike travel speed for various grades. ... 31 Figure 14: Summary of life-cycle GHG emissions from referenced studies. Source is listed on left axis, and emissions reported per-passenger-kilometre travelled [14], [39]– [44]. ... 36 Figure 15: Primary energy source emission intensity as delivered to the bicycle/E-bike and rider ... 39 Figure 16: Primary energy intensity comparison of bicycles and E-bikes. [52], [53] ... 44 Figure 17: Life-cycle Emission Scenario comparisons of bicycle and E-bikes... 46 Figure 18: Full range of potential life-cycle emissions for bicycles and E-bikes compared to other modes of urban transport. ... 47 Figure 19: Free body diagram for E-bike and rider as used to develop the model. Image comes from www.sustrans.org.uk ... 53 Figure 20: Chung method iterative output for drag-area coefficient estimation. Top is first guess, bottom is final iteration of method showing CDA and CRR values. ... 56

(8)

Figure 22: Dynamic trip elevation and speed profile. 1 Hz sample rate, no data filtering. ... 61 Figure 23: Model power predictions for simple trip. Top shows prediction for recorded grade, bottom shows prediction for grade artifically entered as zero for simple trip. Right side shows distribution of predicted power error. ... 63 Figure 24: Example of discrepancy between recorded power and predicted power for dynamic trip behaviour between 1095 and 1110 seconds. ... 66 Figure 25: Human power contributions using Equation 5.1 and CRD trip data for a variety of assist factors... 69 Figure 26: Human power contributions using Equation 5.1 and CRD trip data for a variety of assist factors with impact of grade removed. ... 70 Figure 27: Surface fit and scattered data of all recorded CRD trip power plotted against the speed of the E-bike and the decimal % grade of the roadway. ... 71 Figure 28: Typical efficiency savings when using switching method compared to baseline for a representative duty cycle. ... 72 Figure 29: Histogram of relative efficiency savings per-trip provided by switching

method relative to baseline. Each figure shows different assist factor. ... 73 Figure 30: Absolute efficiency savings per-trip between baseline and switching methods for differing assist factors. Trips sorted by length. ... 74 Figure 31: Histogram of mean motor-power per trip for all trips and for varied cargo load. Each count represents one trip from the CRD data. Red bars represent nominal motor power limit of 500 W according to BC regulations. ... 80 Figure 32:Histogram of instantaneous motor-power for all data points at various loads. Each count represents one second of data from the CRD data set. Red bars represent nominal motor power limit of 500 W according to BC regulations. ... 81 Figure 33:Time-series human and motor power demands from model for an individual trip recorded from CRD data. ... 82

(9)

Acknowledgments

I would like to thank:

My wife for her boundless love and support through this long and time-consuming process. Dr. Ned Djilali, and Dr. Curran Crawford, for their unique insights, experience, and patience. Pacific Institute for Climate Solutions for funding my research.

and last but not least, Coffee, for without your support I could never make it through the endless nights and long days.

We live in this culture of endless extraction and disposal: extraction from the earth, extraction from people’s bodies, from communities, as if there’s no limit, as if there’s no consequence to how we’re taking and

disposing, and as if it can go on endlessly. We are reaching the breaking point on multiple levels. Communities are breaking, the planet is breaking, people’s bodies are breaking. We are taking too

much.

(10)

1. Introduction

Transportation systems, particularly in urban environments, are having to be adapted to increasingly restrictive constraints. From mandatory reductions in GHG emissions, to increased concern about land requirements, infrastructure degradation, and human health impacts, transportation systems are having to change, and personal conveyance choices are one of the primary aspects through which these problems can be solved. While electric cars can reduce GHG emissions, they still as require a high level of energy per passenger kilometre as well as exacerbating land use constraints in dense urban

environments. Improved mass transit solutions can reduce land constraints and address energy use, but often still cause significant infrastructure degradation over time due to large vehicle mass and can have large costs associated with them.

All of the major modes of urban transport struggle with one or more of these

challenges. As an alternative, E-bikes represent a personal conveyance technology that can tackle many of these issues at once in urban environments: they have small space requirements per user, typically have very small operational GHG emissions, and have a negligible contribution to infrastructure degradation.

Currently, E-bikes for personal transport are considered cost prohibitive when compared to traditional bicycles. Entry level E-bikes are an order of magnitude more expensive than entry level bicycles. Cost is only part of the underlying cause of a low adoption rate for E-bikes in North America; the perceived value of E-bikes prior to using them is also a barrier. It appears that there is confusion among the general population as to the value inherent in E-bikes because of a lack of understanding as to their physical capabilities and what their optimal role is in transportation systems.

Additionally, from a commercial perspective, the lack of well-defined capabilities of E-bikes for cargo delivery is also problematic. Fleet operators need well defined metrics

(11)

showing the utility of a particular vehicle (cargo capacity, energy requirements, pace, etc) in order to integrate it successfully into a well managed commercial fleet.

From a government and policy perspective, the full environmental life-cycle costs of E-bikes are unclear. While some Life Cycle Analysis (LCA) work has been done to date, none has explored the upstream emissions associated with human powered mechanical work that would be of concern to a national level government whose domain covers a vast array of GHG emission sources.

This lack of knowledge surrounding the capabilities and costs of E-bikes for a variety of roles is holding back their greater adoption as an environmentally and logistically effective mode of transport in urban environments. As with many other technologies used for transportation, understanding the demands placed on the technology with higher fidelity will allow for more intelligent design decisions to be made, thus improving their performance and reducing their cost, and inform policy developments that can take advantage of evolving technology.

The work in this thesis forms part of a larger project funded by the Pacific Institute for Climate Solutions (PICS) investigating solutions to transportation-based climate issues in the province of British Columbia.

The remainder of the introductory section is intended as a primer on 'E-bike' technology, along with a discussion of relevant problems facing urban transportation systems, followed by an overview of the current state of research surrounding 'E-bikes'. The introduction section ends with an explanation of the specific objectives and

contributions, and an overview of the structure of the thesis. 1.1. E-bike Technology Overview

E-bikes as considered in this thesis fall into a category that is most common among Western markets [1]. Built upon a traditional style bicycle frame, with none of the plastic

(12)

fairing that is typically included with Chinese scooter-style E-bikes. Two typical configurations are shown in Figure 1. On the left is a mid-drive style E-bike where the motor is integrated into the bottom bracket of the bicycle (the pedal location), and on the right is a hub-drive E-bike with the motor integrated into the hub of one of the wheels. Most common among current large brands in western markets are E-bikes operated only in a pedal-assist or power-assist mode where the motor-system only supplies power while the rider pedals. Throttle style E-bikes can still be purchased but they are becoming less common due to regulatory limitations in the European Union that forbid the use of independent throttle E-bikes on public paths and roadways.

Figure 1: On the left, a typical Mid-Drive, geared motor E-bike, the Norco VLT as used in CRD E-bike Trial. On the right, a hub-located direct-drive e-bike. Images from

www.norco.com and www.publicbikes.com

In addition to the position, E-bike motors can also be internally geared or direct drive. Geared motors are more common among large commercial brands but both types are still readily available. Direct drive tends be heavier relative to the nominal power rating when compared to a geared motor. This is partially due to direct drive motors being hub

mounted, and that the rotational speeds for the motor are relatively low from an electric motor standpoint (approximately 250 RPM). Direct drive hub-motors are always

mechanically engaged, which means that they can benefit from regenerative braking, but also that the internal resistance of the stator and rotor must be overcome when pedalling without assist (although this rarely happens). Geared motors contrast the direct drive in that they tend to be smaller, spin at higher RPMs due to internal gearing, and often have more efficient torque output than a comparable direct drive motor.

(13)

Historically, E-bikes started mostly as an enthusiast project where conversion kits would be used on standard bicycles. They typically took the form of rear-wheel or front-wheel hub kits along with a battery mounted on the rear rack of the bicycle. OEM E-bikes didn't gain market relevance until the past decade and only started dominating the E-bike market in the past few years. With this shift, the retrofit market has shrunk

significantly, with one of the major suppliers of retrofit kits going into receivership in the past year [2]. Fully integrated E-bikes are now the dominant form. These fully integrated E-bikes have also supported the use of frame-integrated mid-drive motors (which require a custom frame to fit the motor in place of a typical bicycle bottom-bracket).

Practically every E-bike sold in western markets uses lithium battery technology, which offers the best balance of lifetime cycle count, energy density, and cost for the application of E-bikes. While the exact chemistry of each brand's battery pack isn't easily distinguished, the primary two chemistries appear to be lithium iron-phosphate, and lithium cobalt manganese. The benefits of one chemistry compared to another are mostly due to differences in energy density, current limits, safety, cost, and life time cycle limits.

Aside from the variation in physical design of E-bikes, there is also a variety of regulatory constraints placed on the use of bikes on public roadways. In Canada, E-bikes are regulated through the Canada Motor Vehicle Safety Regulations (MVSR) to have no more than a 500 watt nominal power rating on the motor, to supply no electric assist past 32 kph, and to have fully operable pedals. E-bikes fitting within these

constraints do not require a license or registration to be operated (similar to a bicycle). In Europe, the motor power is restricted to less than 250 Watts, and a maximum speed while under assistive motor power of 25 kph. There are some region-specific variations in Europe. Denmark allows 'speed' pedal assist E-bikes able to achieve motor-assisted speeds of up to 45 km/hr but only on designated cycle paths. In the United States of America, federal regulations limit E-bike motors to no more than 750 watts of power, and a top speed when assisted by the motor of 32 km/hr. As long as the E-bike meets these regulations, no registration, insurance, or driver's license is required to use one on public

(14)

trails and roadways. There are E-bikes on the market not subject to these regulations, but they are restricted to off-road applications such as mountain biking.

1.2. Literature Review

This section provides an overview of the state of E-bike use, along with the current state of E-bike related research. Background information on issues facing urban transportation systems (specifically automobile dominated urban transportation), is followed by a review of research progress. Later sections will provide more targeted literature reviews addressing the subject at hand (i.e. modelling, GHG emissions, etc.). The following areas are outside the scope in this thesis:

• human physiological response to cycling and e-bikes, • human psychology of e-bike and bicycle use,

• civil infrastructure considerations, • detailed electric motor design,

• detailed battery chemistry or mechanical e-bike configurations.

1.2.1. Driving Factors for Change in Urban Transportation

Urban transportation networks are under an unprecedented set of challenges with a wide variety of causes. Land constraints, GHG emissions, air quality, and energy limits are causing the way society looks at transportation in urban environments to change. How these challenges will be dealt with over the coming decade is still to be determined but it is likely to be through greater emphasis on multi-modal transportation system design. In order to understand how E-bikes can be a solution, a better understanding of how other personal conveyance choices contribute to these problems is required.

Land constraints are a major driving force in urban environments. With the share of urban populations nearly doubling over the last 50 years [3], the demand placed on each square metre has increased. Many North American cities have responded to this increased demand for space by encouraging urban sprawl through the development of sub-urban zones. One major impact is an increasing reliance on cars to perform all trips; the farther people live from urban centres, the more time they spend travelling in personal

(15)

with some major urban centres, such as Tokyo, New York, and Paris, having as much as 25% of their total urban land dedicated to roadways [5].

As urban populations continue to grow, and with cities running low on available land for development, the various transportation options available for use in urban

environments can play a large role in either exacerbating this problem or offering relief. Table 1 shows some estimates of the land requirements of various modes of transport in urban environments. According to these estimates, walking, bicycling and bus transit provide the most space efficient forms of transport, but like most things, there is more to consider than just land requirements.

Table 1: Estimated land area requirements of various modes of transport [6].

Mode Average Speed [kph] Moving Area [m2_] Parking Area [m2_] Total [m2_] Walking 4.8 1 - 1 Bicycle 16 6 3 9 Motorcycle 48 67 14 81 Bus Transit 32 5 - 5

Solo Driving - Urban 48 67 28 95

Solo Driving - Highway

96 195 28 223

Climate altering emissions are a primary factor when comparing urban transportation technologies. As of 2014, transportation was responsible for as much as 14% of global CO2 emissions [7], which increased to 15% by 2017 [8]. In British Columbia in

particular, smart decisions about which transportation technology to support can result in significant reductions. Ambitious targets have been set by Governments around the world to reduce transportation based emissions, with many focused on increasing consumption of renewable fuels and supporting the adoption of electric cars [9], [10]. Along with direct transportation-based emissions, supporting specific technology choices can have indirect impacts on emissions as well. The development of alternative urban

(16)

transportation infrastructure (bike lanes, rapid light rail, etc.) fosters increased urban density, thus reducing total per-capita energy and emission intensities [11].

Fossil fuel-based transportation is a major contributor of climate changing emissions, but electric cars can also have a significant contribution if their electrical energy supply is fossil fuel based. Electric cars address tail-pipe emissions but not necessarily well-to-wheel (WTW) emissions associated with the production of the fuel or energy required to move the vehicle. For countries relying heavily on fossil fuels for electricity generation, electric vehicles have the potential to increase total GHG emissions when replacing fossil fuel-based cars, contrary to initial expectations [12]. The choice of energy carrier

(hydrogen, fossil fuel, biofuel, etc) and battery production are found to be the primary drivers of well-to-wheel emissions in vehicles, therefore choosing to support vehicles that use renewable sources of energy can make dramatic reductions in associated well-to-wheel emissions [13]. Although well-to-well-to-wheel emissions can be reduced with renewable energy sources, the use of electric cars, whether hybrid drive-trains or fully electric, have significantly increased GHG emissions associated with their production when compared to traditional ICE cars [14], [15].

Urban air quality can also have a significant impact on the health of city-dwellers with fossil fuel vehicles responsible for 25% of global urban ambient air pollution [16]. Within large North American cities, 30-45% of the population lives within areas considered highly affected by traffic emissions [17]. Urban air quality concerns are driving change through targeted improvements in transportation infrastructure, improved fuel quality, and alternative transportation technologies [18].

Though alternative fuel cars (biogas, natural gas, electric) can reduce climate changing emissions and urban air pollution, they don't necessarily address the issue of energy intensity of transport. A 1200kg vehicle, whether electric or fossil fuel powered, still requires large amounts of energy to transport a single occupant. Table 2. shows the energy intensity per passenger kilometre of several modes of transportation with occupant values typical for Dublin, Ireland, where the study was based.

(17)

Table 2: Embodied Energy of various transport modes divided among passenger load [19] Mode Embodied Energy [MJ/km] Occupancy Embodied Energy [MJ/pkm] Bicycle 0.11 1 0.11 City Bus 1.37 25 0.05 Private Car 0.73 1.4 0.52 SUV 1.44 1.4 1.03 Light Rail 5.92 428 0.01

As electrification is often touted as a solution for replacing most consumption of fossil fuels (building heating and cooling, transportation, industrial processes, etc.), the demand placed on the future electrical grid is going to be enormous. Heating alone is predicted to require almost a 30% increase in future grid capacity in California, a relatively warm location [20]. From a global perspective, transportation is responsible for nearly 30% of total global energy consumption, of which approximately 92% is fossil fuel based [21]. Electrification of transportation will place enormous demands on National grid

infrastructure, with the UK predicting that it may require up to a 30% increase in energy generation to handle the electrification of its entire fleet [22].

A shift from cars to E-bikes as a major urban transportation mode would result in a dramatic reduction in land-use requirements, GHG emissions, negative urban air quality impacts, and energy use. Some of these impacts have been researched previously (land-use requirements, urban air quality impacts) but others such as GHG emissions and energy use are not as clearly understood in academic literature. The next section explores the existing literature surrounding E-bike use in Western societies.

1.2.2. E-bike Research

E-bike research to-date has covered a wide range of topics, with a particular focus on safety, behaviours, demographics, and environmental performance. While there are many engineering-based papers focusing on the electrical sub-systems of E-bikes, they are

(18)

considered outside the scope of this thesis. A large number of papers focusing on E-bike issues in China are not considered within this thesis for two primary reasons: a large portion of Chinese E-bikes are of the scooter style with large plastic fairings and lead-acid batteries [23], and this thesis focuses primarily on western issues facing E-bikes. The research discussed below is intended to provide the reader with context regarding the current use of E-bikes in western society while highlighting the lack of information on the energy use, emissions, and physical capabilities of E-bikes for transport.

Age and female gender appear to be negatively associated with perceived safety while riding an E-bike when compared to a regular bicycle in Denmark [24]. A common perception of the cause of accidents is other road-users underestimating the speed of E-bikes [24]. Perceptions are different than actual risk, as another study found no

correlation between age and actual accident rates but did find that elderly riders and women were more likely to be severely injured when crashing [25]. While E-bikes tend to travel faster than regular bicycles, another study showed that there is no significant difference in the overall traffic conflict risk between bicycles and E-bikes, although according to this research, E-bikes have a dramatically higher risk of accident at intersections due to the increased average speed [26].

A study out of the Netherlands showed no significant difference between riders of E-bikes and riders of bicycles with respect to rider safety behaviour; E-E-bikes most often travelled at faster speeds with the exception of E-bikes travelling slower than bicycles on shared pathways, and traffic safety violations were comparable between e-bike and bicycle riders [27]. While traffic safety violation rates in the Netherlands are quite similar for both bicycle and E-bike riders, other research has shown that typical usage cases for the two vehicles can differ. E-bikes are more often used for running errands and

commuting when compared to regular bicycles [28]. The same research also shows that the reason for using an E-bike differed between young and old, with Generation X and Millennials choosing E-bikes to save time and reduce environmental impacts, while most other people chose it to increase health outcomes [28], [29]. Typically, as people age,

(19)

they cycle less frequently but E-bikes have been shown to reverse this trend, as well as increasing self-reported cycling distances when compared to regular bicycles [30]–[32]

A review of many studies found that most of the increase in E-bike use comes at the cost of decreased bicycle use, and while the exact environmental cost of this switch isn't known in the literature, E-bike adoption still has noticeable impacts in reducing car use [33]. While the reduction of bicycle use isn't ideal, once switched, E-bike owners use cars noticeably less frequently than bicycle users [33]. The exact nature of mode substitution is very context specific and varies from region to region depending on technology

availability and infrastructure support [33]. Research has shown that in North America at least, most e-bike users rode a traditional bicycle prior to using an e-bike [32]. A

comprehensive study out of the United States in 2014 showed that the majority of E-bike users were male (85%) and white (90%) [32].

Research shows E-bikes safety metrics are similar to bicycles. If the infrastructure is in place, accidents are rare. Current E-bike users tend to be male, college educated, and white; and the reasons they use E-bikes are either altruistic in their attempts to address environmental problems or motivated by health benefits of bicycles while physical capabilities diminish with age. With this understanding, the objectives and contributions of this thesis are presented.

1.3. Objectives and Contributions

This thesis addresses the following question: "How does electric assist alter the environmental impact and physical performance of bicycles, and what are the optimal roles for electric assist bicycles?"

The primary research question will be answered through the following specific objectives:

• Create and validate a high-fidelity energy-based bicycle/E-bike model; • Characterise urban bicycle/E-bike trips with respect to the demands of

(20)

• Analyze contributions of human supplied mechanical work and electrically supplied work to the motion of E-bikes;

• Quantify GHG emissions associated with human supplied mechanical work and compare with emissions from electrically supplied work;

• Quantify physical response of E-bikes + Rider with respect to variation in loading, human power contributions, motor power contributions, and geographic topology.

This work is intended to provide a comprehensive overview of the capabilities and impacts of E-bikes in a variety of scenarios. The primary contributions are to provide a quantitative assessment to inform decisions regarding the applicability of E-bikes for commercial fleets and personal use. With a more detailed understanding of the relationship between human power, electric power, and trip characteristics, further progress can be made with respect to motor design, control systems, commercial fleet deployment, and policy decisions.

1.4. Thesis Structure

The remainder of the work contained in this thesis consists of four primary projects that are detailed in chapters 2 through 4, and are then combined for several sets of analysis in Chapter 5. The content of each chapter is detailed below:

Chapter 2 covers the experimental data collection campaigns that represent the original data used in this thesis and represents the preliminary results. The two experimental campaigns are presented as subsections, detailing the methodology, analysis, and preliminary results. The first subsection is an analysis of an electric motor with two internal wiring configurations offering distinct performance profiles. The second is an E-bike trial involving 17 participants and several months of urban E-bike trip data.

Chapter 3 shows the investigation of the life-cycle environmental performance of E-bikes and bicycles relative to other primary modes of urban transportation. This chapter also includes analysis to quantify the upstream emissions that occur from accounting for

(21)

the food-based energy supplied to produce human-mechanical work for E-bikes and bicycles.

Chapter 4 covers the development of a mathematical model to predict the energy use that occurs while riding a bicycle and an E-bike. A brief literature review is presented followed by the derivation and validation of the model.

Chapter 5 documents the several different sets of analysis performed with the experimental data, the environmental data from the LCA, and the mathematical model. This includes an assessment of the human power contributions during the CRD trial, an performance assessment of a novel electric motor configuration, quantification of the performance of E-bikes deployed in a municipal urban fleet, and finally an investigation into the energy demands of E-bikes used for urban cargo delivery.

Chapter 6 offers a discussion of the results in this thesis, final conclusions that can be drawn from this work, as well as recommendations for future work.

(22)

2. Experimental Data Collection

This section of the thesis summarises the two major data collection campaigns conducted as part of this research. The first comprises a series of E-bike electric motor performance characterisation tests captured in a laboratory setting. The second captures trip characteristics and human riding behaviour through the deployment of E-bikes in a commercial fleet. This chapter summarises the data collection methods, the data, some preliminary analysis and summary results. In-depth analysis and predictive modelling using the data presented here occurs in Chapters 3 and 5.

2.1. Exro Motor Performance Characterisation

The Exro Project was a partnership between myself, my academic supervisors Dr.’s Ned Djilali and Curran Crawford, and Exro Technologies. The project was funded with a National Science and Engineering Research Council (NSERC) Engage grant designed to foster relationships between academia and industry such that Canadian based innovative research can be improved. The purpose of the research from Exro’s perspective was to quantify the effect of Exro’s switching technology as applied to E-bikes. From my perspective, I had the added goal of obtaining empirical data detailing the efficiency of a commercial E-bike motor within the range of typical urban-use duty-cycles.

Two approaches are used to achieve these goals: the first is to quantify the performance of a typical commercial E-bike motor for both its off-the-shelf operation and with Exro’s switching technology, and the second is to cross compare this laboratory performance with real-world urban E-bike duty cycle data from the CRD project (detailed in section 2.2).

2.1.1. Experimental Set-up

Multiple experiment configurations were used over the course of the project, with any opportunism for improvement in data collection and analysis applied. The initial

(23)

Table 3: Summary of Initial Exro Project Test Equipment

Item Purpose Notes

Miele Veneto GR2 Bicycle In-situ test bed for electric motor

-

Bionx P350 Motor Typical commercial E-bike motor

Direct drive, rear-hub mounted, 350 W nominal power, 32 kph speed governor

Bionx Battery Power for motor 48V, 317 Wh

Grin Cycle Analyst V3 Power supply and CPU for data logger

Connected with Shunt between Bionx motor and battery

Grin Analogger Data logger for motor and energy consumption

-

Wahoo Kickr Smart Trainer

Rear-wheel power meter measure total bicycle power output

On-wheel power meter with eddy current brake

The BionX P350 motor used in the initial tests was a stock unit with no modifications and was connected to a 48V lithium-based battery. The Cycle Analyst (CA) and the Analogger were used to automatically record the voltage and current between the battery and motor.

The Wahoo Kickr Smart Trainer monitored the power output of the E-bike through an eddy-current braking system. The braking loads was controlled through a mobile phone application that allows the user to set a percentage of the total available braking force. The Smart Trainer then uses a proprietary model-based power metering method to estimate the power output; the Smart Trainer is stated as having a ±3 accuracy but that the accuracy improves once the system has warmed up.

(24)

The second round of tests had a modified equipment set-up that accounted for issues that arose during the first round of testing. In addition to improving the testing methods, the Bionx P350 was opened and rewired internally to allow for the coils to be run either in a parallel configuration (same as stock) or a series configuration. The rewiring, completed by Exro staff, required the removal of the Bionx speed controller, which was housed inside the Bionx motor casing, in order to accommodate the space required by the additional motor wires. An external third party speed controller was used in place of the Bionx controller. Additionally, a dedicated DC power supply was used to remove and state of charge (SOC) issues. Table 4 shows the equipment used for the second round of testing.

Table 4: Summary of Final Expo Project Test Equipment

Item Purpose Notes

Toba Edison Bicycle In-situ test bed for electric motor

-

Bionx P350 Motor Typical commercial E-bike motor

Direct drive, rear-hub mounted, 350 watt nominal power, rewired and speed control removed

Volteq HY502EX DC Power for motor 50V, 20A max

Fluke 289 Multimeter Monitor DC supply voltage - Fluke 325 Clap meter Monitor DC supply current - Wahoo Kickr Smart

Trainer

Rear-wheel power meter measure total bicycle power output

On-wheel power meter with eddy current brake

Speed controller RPM based motor controller

Restricts maximum current and power demands of the motor relative to first round testing.

(25)

2.1.2. First Testing Method

The first testing method used the equipment listed in Table 3 to generate a series of RPM and torque states for the Bionx motor and bicycle system. The states covered the range of typical bicycle duty cycles, with up to approximately 270 RPM (32 kph) and up to approximately 30 N of torque. At each recorded state, the power input and output of the motor was measured, and the efficiency of the E-bike system was calculated. The efficiency values included the losses across the motor, through the contact between the wheel and the roller based smart trainer (vibration, friction, contact resistance), and losses due to the inertia of the bicycle wheel and the smart trainer flywheel. All data points were not steady state during this testing method.

The first testing method consisted of mounting the E-bike on the smart trainer, setting the smart trainer eddy current brake resistance, and applying a throttle signal to the Bionx system to steadily accelerate to the maximum speed (32 kph) under load. This process was then repeated for successively larger braking loads (increased eddy current brake resistance) until the maximum achievable torque output was reached.

Figure 2 shows the response of the system to the human controlled throttle input for an individual test and is intended to show the variability in the system response to the human controlled throttle. Since the power was recorded in a dynamic state, the variations in sample rates between the input power monitoring and the output power monitoring caused negative impacts on the fidelity of the resulting calculations. An increase in power input did not always temporally match the power output.

(26)

Figure 2: Example of initial test results for a fixed braking force as the Bionx system is accelerated from rest to maximum speed.

Figure 2 shows in a few spots that an increase in power input was followed several seconds later by an increase in power output. This temporal offset caused issues in calculating efficiency at each time step in the recorded. Also, since the throttle control was very sensitive, it was difficult to achieve slow and steady accelerations to minimize the temporal effects. This could not be remedied by simply shifting the data sets, as the temporal discrepancies were not consistent throughout a given test.

The input power data (recorded as voltage and current by the CA shunted between the Bionx battery and Bionx motor), and the output power data (recorded as a single power metric from the smart trainer) were in two different data files with different sample rates and different relative time-stamps. An attempted remedy was to align the two signals using the point of maximum cross-correlation between the two signals1.

1_{MATLAB’s built in ‘xcorr.m’ function was used to measure the similarity of the two signals with the} maximum of the output of ‘xcorr.m’ being used as a delay to align the two signals.

(27)

The individual test of Figure 2 was repeated for the whole range of available braking forces. The recorded power output was converted to a torque value using equation 1, with the resulting torque-RPM state data for two full rounds of testing shown in Figure 3. At each state, the current and voltage input to the motor was recorded, as well as the speed and power output of the motor. This entire process was repeated several times to determine whether the results were consistent across multiple trials using the same method.

𝜏 = 𝑃𝑜𝑤𝑒𝑟 2 𝜋𝑅𝑃𝑀₆₀

(1)

Figure 3: Entire range of achieved torque-RPM states for two rounds of testing using the first method and equipment. A) first round of testing, b) second round of testing

Efficiency maps using the recorded state data are shown in Figure 4. The maps were generated by applying a polynomial fit2 to the scattered data which consists of the hundreds of torque-RPM states recorded during testing. The efficiency at each recorded state was calculated using the ratio of power input from the battery to the motor, and power output of the motor at the wheel-cycle trainer interface. These two surface fits show fairly significant differences in the estimated efficiency for a given torque-RPM state.

2_{The polynomial fit was generated using the MATLAB function ‘fit.m’ with the ‘poly32’ fit type option, and} the experimental data input as a scattered data set.

(28)

Figure 4: Polynomial surface fits to A) round 1 and B) round 2 state data of first testing method. Contours show system efficiency.

The absolute difference in efficiency between the two rounds of testing is shown in Figure 5. Up to a 20% difference can be seen between the two surfaces. This is especially prevalent in what will later be shown to be the primary operating states for typical urban duty-cycles.

Figure 5: Absolute difference between the two efficiency surface fits of Figure 4

A brief analysis of the surface fit was done to ensure that fitting was not one of the major causes of the discrepancies between the multiple rounds of testing. The residuals of

(29)

the surface fit for both rounds of testing are shown in Figure 6 along with an assessment of how well the error is distributed. One metric of a good fit is to show a random

distribution of the error centred around zero, which can be seen in both cases of Figure 6. Additional goodness of fit metrics for each surface are shown in Table 5: the RMSE for both is relatively small compared to the scale of the dependent variable (efficiency from 0% to 100%); the R-squared value is not great, showing that the modeled surface fit only accounts for 79% and 60% of the variability of the experimental data for round 1 and 2 respectively. Other surface fit types were using within the MATLAB toolbox with the method used in this thesis found to be the most accurate.

Figure 6: Distribution of residuals for surface fit to a) round 1 and b) round 2 data using the first testing method. Mean and standard deviation of error from surface fit shown for

each round of testing.

Table 5: Quality of fit parameters for surface fit to first testing method data.

Parameter Round 1 Round 2

SSE 2119 3584

(30)

R Square 0.79 0.60

Ultimately, it was felt that the test results were too heavily impacted by the battery SOC, the variability in the temperature of the smart trainer eddy current brake, the sensitivity of the throttle signal, and the apparent temporal difference between recorded powers. Attempts were made to minimize these issues by starting with a full SOC and working through the tests in the same order for each round but due to the difficulty of controlling the throttle signal, the SOC was not the same from one round to the next. Significant changes were made to the experimental set-up in an attempt to counter these issues, which form the basis of the next section.

2.1.3. Second Testing Method

The second testing method used the equipment shown in Table 4. A third-party RPM based speed controller was used to remove any of the human throttle control issues, and the DC power supply was used to remove any of the SOC issues. Before data was collected, the smart trainer was warmed up by setting a braking load and allowing the system to run for 20 minutes in an attempt to reach a relatively stable thermal state. All further tests were run by setting a fixed braking load with the smart trainer, setting a target RPM, allowing the system to reach a steady state, and then recording all of the data points at the steady state (power supply voltage, power supply current, motor power output, and motor RPM). This process was repeated for the entire range of achievable RPM and for successively increasing braking loads.

For the second testing method, rather than capturing a continuous run of data as in the first method, steady state data was captured at a range of states that span as similar a domain as possible to real-world duty-cycle states. This second method removed the temporal misalignment between input and output power that was present in the firs testing method. The only remaining obstacle was that the speed controller used in the second testing method had internal electrical limits that restricted the upper range of torque values achievable. The states obtained using the second testing method are shown in Figure 7.

(31)

Figure 7: Achieved torque-RPM states for two rounds of testing using the second method for both parallel and series wiring configurations. a) and b) are parallel tests, c) and d) are

series tests.

The surface fits for each round of data for both series and parallel configurations are shown in Figure 83. The surface fitting function in MATLAB extends the surface to the maximum of the domain along both axes regardless of whether the experimental data covers the entire range. This means that a significant portion of the series surface fits of Figure 8 extend beyond the range of experimental data. For both the series and parallel results, the surfaces do not extend below approximately 50 RPM as in most cases the cycle trainer returned a value of zero power due to the limitations of the cycle trainer.

(32)

Figure 8: Polynomial surface fits to second method results. a) and b) use parallel test data, c) and d) use series test data.

Significant differences can be seen between the efficiency surfaces of the series and parallel configuration, but this is as expected. The significance of these differences is explored further in Chapter 5. The absolute differences between each round for both the parallel and series configurations are shown in Figure 9 with the results showing much better agreement than with the first testing method. The differences for this second testing method peak at approximately 12% at the upper limit of experimental data (15 Nm) with no more than 4-5% for any speed with less than 12-13 Nm of torque. For the series tests, the more accurate testing region was high torque and low RPM, with less than 4% difference for almost all values less than 18 Nm of torque and less than 200 RPM.

(33)

Figure 9: Absolute difference between round and 2 efficiency maps for second testing method. a) parallel and b) series wiring configurations.

The quality of fit for each surface is examined to understand what contributions they might have on the error between tests. As in the previous section, the distribution of the residuals and the other quality of fit metrics are presented in Figure 10 and Table 6.

Table 6: Quality of fit parameters for surface fit to second testing method data.

Parameter Parallel Round 1 Parallel Round 2 Series Round 1 Series Round 2 SSE 795 2071 261 840 RMSE 3.47 5.65 2.66 4.7 R Square 0.94 0.92 0.93 0.84

(34)

Figure 10: Distribution of residuals for surface fit to round 1 and 2 results using second testing method. a) Parallel round 1, b) parallel round 2, c) series round 1, d) series round 2.

The second testing method produces significantly smaller SSE and RMSE as well as an increased R-squared value when compared to the results of the first testing method in Table 5. This is very likely due to the steady state nature of the recorded data. By

recording data at steady state, the temporal noise discussed in the previous section, along with the throttle sensitivity issues, were effectively removed. The removal of the SOC impacts improved the repeatability and reliability of the tests, although further analysis using the second method efficiency maps of Figure 8 requires the caveat that it assumes all operation is at full battery SOC.

(35)

2.1.4. Exro Project Summary Results

The tests for each configuration are combined to generate a final efficiency map for both the series and parallel configuration. These efficiency maps are presented in. These efficiency maps were generated by combining the raw data points from both rounds of testing prior to surface fitting. Figure 11 shows the clear difference in performance offered by the series and parallel wiring configurations. The series configuration offers better efficiency at medium to high levels of torque and low RPM while the parallel configuration contrasts this with higher efficiency at high RPM and low to medium torque values.

Figure 11: Efficiency maps for both rounds combined. a) parallel combined efficiency map, b) series combined efficiency map, c) parallel individual efficiency data points, d) series individual efficiency data points.

(36)

2.2. CRD E-bike Trial

The CRD Project is a joint project between myself and my supervisors, Dr.’s Ned Djilali and Curran Crawford as members of IESVIC, and the Capital Regional District (CRD), that is funded by a grant from the Green Municipal Fund. The CRD is a local governmental organization that oversees the region-wide management of services that can be shared between the thirteen municipalities of Greater Victoria. The E-bike project forms one part of a larger FCM funded transportation program that focuses on reducing the transportation-based emissions of the CRD while incorporating academic research.

The specific goals of the CRD E-bike Trial are:

• To characterize the energy use and physical capabilities of E-bikes for urban trips • To quantify the reduction in GHG emissions of the CRD fleet through the

substitution of E-bikes for car trips.

The results of the pursuing the first goal are reported in this Chapter, while the substitution impacts are reported in Chapter 5.

2.2.1. Experimental Set-up

Both goals were achieved through the deployment of three E-bikes outfitted with sensors that logged performance metrics during each trip. The sensor package installed on each E-bike consisted of a Garmin Edge 520 cycle computer, a Garmin ANT+ protocol speed sensor mounted on the front wheel hub, and a PowerTap Ant+ protocol hub-based power meter built into the rear wheel. Due to the location of the power meter, only total power output of the E-bike with no explicit differentiation between the rider and the electric motor is recorded. Later analysis will break down the results into human and motor contributions.

Each of these sensors was connected to a Norco VLT R1 E-bike, synced to the Garmin Edge 520, with the data collected from the Edge on a weekly basis. The CRD staff involved in the project could reserve and E-bike through the CRD’s internal online vehicle booking system. Each time they rode the E-bike, they would simply press a button on the Edge to initiate data logging and press the same button to end the ride and

(37)

save the data. Seventeen users were recruited into the project, with each rider’s trip data anonymized to meet CRD privacy concerns.

By the end of data collection, the CRD project resulted in a large number of trips representing over 4 months of data. There was a significant amount of non-compliance when it came to recording data. The E-bike odometers showed a total of nearly 1200 km with the actively recorded data, summarized in Table 7 showing only approximately 600 km. While data logging was optional for CRD staff, this missed data did likely impact the fidelity of the results as many trips were missed.

Table 7: CRD Project data collection summary.

Metric Value

Total kilometres travelled 607 km

Number of trips 92

Average speed 20.3 ± 5.9 kph

Average trip length 6.6 ± 5.8 km

Average trip time 25.9 ± 25.2 min

By capturing both speed and grade, the CRD data can then be assumed to represent an estimate of typical urban trips, although with comparative data from other regions or fleets, it is difficult to say how transferrable the results are to other jurisdictions.

Conversely, since the E-bike speed limiter is almost universal among e-bikes in Canada as governing speed to a maximum of 32 kph, and the motor power rating is also

regulated, it can be assumed that the typical speed profiles would have a relatively consistent

2.2.2. CRD Summary

The energy values were recorded using the PowerTap G3 power meter and are reported in two different forms in this thesis: either as primary energy use (dietary and electrical) or as the total energy use required to overcome air resistance, rolling resistance, and to make the mass of the E-bike+rider system accelerate. In this chapter, only the total

(38)

energy use is reported, with primary energy use detailed in Chapter 3. The power data, along with the other ride characteristics (speed, location, grade) are used to understand how and when energy was expended during the trip. The energy use also allows for determination of the GHG emissions that occur from using the E-bike, as detailed in chapters 3 and 5.

Table 8 shows the average energy of the total recorded data. In addition to the summary table, the recorded power data can be categorized into total energy use in response to distance, grade, and speed to provide further insight as to how E-bikes are used in an urban setting.

Table 8: Average energy and power for total recorded data

Metric Value

Average per-trip power 234 ± 73 W

Average per-trip energy use 40.3 ± 48.0 Wh Average per-kilometre energy use 7.8 ± 2.5 Wh/km

The variation between trips is documented in Figure 12, which shows per-trip values for total energy use, the average instantaneous power, and the distance per-trip. Some of the trips didn’t have any recorded power values due to issues with the sensors pairing improperly during use by the CRD staff; this power-less data was removed from the summary in Figure 12.

(39)

Figure 12: Per-trip energy use (top), average per-trip power (middle), and trip distance (bottom), as recorded during the CRD project.

(40)

Figure 13: E-bike travel speed for various grades.

2.3. Experimental Conclusions

The collected data provides insights on how E-bikes are ridden in urban environments along with detailed performance characterisations of a typical E-bike motor. Several assumptions are supported by the experimental data. Series and parallel winding configurations offer distinct optimal operating regimes with respect to RPM-torque states. In addition, while the final Exro results showed some inconsistency between rounds of testing, for the primary operating range (shown in Chapter 5), the absolute efficiency differences tend to be less than 4%.

In order to improve upon the current Exro results, several options can be considered. The specific objectives of future work will dictate which choices should be made. As the work in this thesis was aimed at characterising the system efficiency of the E-bike motor in-situ, a power meter that interfaced with the rear-wheel was required. If better

characterisation of the motor in isolation is desired, then a bench-top dynamometer would likely achieve more accurate and repeatable results. Building a more detailed analytical

(41)

model of the bicycle that accounted for efficiency losses through the wheel could be combined with the bench-top dynamometer results to allow a simulated total E-bike system efficiency.

For the CRD data, we can surmise that from a fleet operations standpoint, users appear to have no problem maintaining total motive power output for trips up to at least 10 kilometres total. Average power across all trips was relatively consistent with the

standard deviation representing only approximately 30% of mean power. When queried, most participants stated they used a variety of assist levels, meaning that even with the assist factor in the data sets changing from 0.5 to 2.75, the average instantaneous power did not vary greatly. From a fleet managers perspective, this data shows that E-bikes can fill a niche without human endurance limiting E-bike success.

There are several avenues to improve upon the findings from the CRD campaign. The first, which was considered and rejected due to the added burden it would place on the CRD participants, was the addition of a second power meter in the pedals that would record the human power output during all trips. This would have required a second cycle computer, and since the user compliance rate for the CRD was relatively low (only approximately 50% of all trips were recorded), this would likely have led to even lower levels of recorded data. Second, a similar trial with a bicycle without electric assist would have given more detailed data about un-assisted human power for comparison purposes later in this thesis. Finally, a comprehensive survey of CRD participants would have given further insight as to a range of factors that would influence the power meter readings: user fitness, experience with bicycles, aversion to rain, purpose of trips, etc. Employer confidentially issues and academic ethics study procedures were too onerous to include a survey of this type during the CRD project.

(42)

3. E-bike Emissions and Energy Use

This chapter is a partial reproduction of a paper in-review with the International Journal of Sustainable Transportation. The reproduction here omits some parts in favour of keeping them in Chapter 2, where they fit better within this thesis. The paper here synthesizes the results of several lifecycle analysis studies, builds a range of dietary and grid emission intensities, and then determines the primary energy-use of E-bikes and bicycles using data from the CRD project. All of these parts are combined to determine what is proposed as a higher fidelity estimate of the potential whole lifecycle emissions that come from the use of bicycles and E-bikes.

This chapter is developed to answer the environmental aspect of the thesis question: How does electric assist augment (or detriment) the environmental performance of bicycles when considering the full life cycle and fuel/energy sources.

3.1. Introduction

As of 2014, fossil fuel based transportation is responsible for 14% of total global CO2

emissions [34]. With these significant emissions and air quality issues facing urban transportation systems the benefits of a shift to cycling would appear obvious. Municipal authorities from around the world are significantly expanding cycling infrastructure in the hopes of stimulating mode switching. Decades of infrastructure investment in locations such as Amsterdam and Copenhagen have helped achieve cycling mode shares in the range of 17-41% [35], [36]. Along with the growth in traditional bicycle mode share, many European cities are also seeing a large growth in the adoption of electric bicycles (bikes) as costs and performance have improved and availability increased [37]. E-bikes as considered in our study, vs. E-scooters, still require human pedal input power but augment that power proportionally with an electric motor.

Survey results of early adopters of E-bikes found that they were more likely to switch from driving to cycling with an E-bike as opposed to a traditional bicycle, and that they made more trips with an E-bike than with a bicycle before having access to the E-bike [38]. An increase in willingness to use a bicycle once it is electrified is attributed to the

(43)

rider's perception of a decrease in the significance of hills and overall exertion required to reach a destination: E-bikes make cycling feasible for a broader population and over a broader range of routes. If E-bikes can address some of the shortcomings of the

traditional bicycle, then what are the challenges facing E-bikes? It is relatively straight forward to surmise that adding an electric motor and battery to a bicycle will increase its environmental impact during the production phase, but an open question is how this might impact the use phase and its relative share of life-cycle emissions. Seldom

acknowledged in the current body of research on environmental performance of bicycles is the fact that their power source is a food-fuelled human body [39]–[41]. Food

production is generally an energy intensive process and its environmental performance when compared to electricity as a method for powering transportation has yet to be thoroughly explored in the realm of bicycles.

The motivation of this paper is to explore the range of possible upstream GHG

emissions due to the caloric intake required to balance the human mechanical work used for riding bicycles and E-bikes, as well as the range of grid emissions associated with powering the electric assist. While this work references several published LCA, and expands upon them with empirical e-bike trip data, it does not offer clarity on the impacts of long term changes in energy intake (increased food consumption) in response to increased in energy expenditure (exercise). An LCA that truly accounts for the impacts of active transportation on increased caloric intake would be quite difficult due to the long time required for equilibrium between energy expenditure and energy intake (on the order of several months) with the introduction of increased exercise. This work is meant to show a broad range of possible emissions scenarios that should catch within its boundaries the reality.

Additionally, this work is intended to show an estimate of the energy required by both the human provided power and that provided by the electric assist. A dearth of

longitudinal physiological studies fully examining the interplay of diet and increased regular exercise, along with the obesity range of the population, requires bounding the caloric input impacts.

(44)

Establishing the range of possible emissions from E-bike and bicycle use can help determine the extent to which dietary and grid-linked emissions bound the relative GHG benefits of bicycles and E-bikes.

A rational approach to account for and compare emissions associated with human power for bicycles and E-bikes is to perform a seed-to-wheel life-cycle analysis that parallels well-to-wheel analyses used for plug-in vehicles. In the first stage of analysis, a review and aggregation of several Life-Cycle Analyses (LCA) studies is performed to create a baseline for several primary urban transportation modes. Building on this baseline, further analysis considers the GHG emissions intensities of both human power and electric power. The human-supplied mechanical work is assessed by accounting for energy expenditures of commuter cyclists from both third-party research and the results of an experimental campaign conducted by the authors, which are then combined with dietary emission intensities of typical Western diets. The electrical power analysis is performed using electrical grid emission intensities corresponding to the location of the cited dietary emission intensities. In order to provide greater context to the E-bike and bicycle emissions results, other primary modes of urban transport are also considered in the comparative analysis: cars, buses, and electric cars (BEV). The GHG emissions associated with buses, cars, and electric cars are aggregated from previous studies along with a brief review of the assumptions and limitations from each study. Finally, the paper concludes with the insights and further questions gleaned from the study.

3.2. Referenced Life-Cycle Emissions

Seven distinct LCA studies that modelled the impacts of bicycles, E-bikes, traditional internal combustion cars, battery electric cars, and buses are used as input to our analysis. These particular studies were chosen because within each study, they covered a range of transportation options and employed self-consistent methodologies. Intra-study

methodology variations were inevitable and were used to generate an associated while capturing a wider range of potential attributable emissions within each mode.

(45)

The results from each of the referenced studies are shown in Figure 14. They are separated into production phase emissions and use-phase emissions (where delineated), otherwise only the total life-cycle GHG emissions were reported. The functional unit of most of the results in this paper are presented as kilograms of carbon dioxide equivalent emissions per-passenger kilometre travelled. This metric accounts for the varying

passenger loads of each mode of transport, variations in assumed functional lifetime, and total kilometres travelled over that lifetime. Within the referenced bicycle and E-bike LCA, the studies did account for embodied emissions associated with infrastructure for the use of the vehicles, as well as the production of batteries, motors, and other electronic components. This puts the E-bike and bicycle LCA boundaries in a similar position to those of buses, battery electric cars, cars.

Figure 14: Summary of life-cycle GHG emissions from referenced studies. Source is listed on left axis, and emissions reported per-passenger-kilometre travelled [14], [39]–[44].

The variation among the bicycle and E-bike results in Figure 14 are primarily due to differences in operational lifetime, average speed (and thus energy use), and battery